search

DAMO-NLP-SG / VideoLLaMA2

VideoLLaMA 2: Advancing Spatial-Temporal Modeling and Audio Understanding in Video-LLMs
Apache License 2.0
752 stars 50 forks source link

Video LLaVA dataset issue #20

Closed ayaan-together closed 3 months ago

ayaan-together commented 3 months ago

In the finetuning phase, I get [h264 @ 0x1f0acd40] mmco: unref short failure and [mov,mp4,m4a,3gp,3g2,mj2 @ 0x19a50700] moov atom not found. It does not stop training though.

Also, could you tell me how many videos there are in valley and videochatgpt datasets. The download I have has many missing files.

lixin4ever commented 3 months ago

Hi Ayaan, good to receive your feedback.

The warning does not indicate the FileNotFound error but some internal issues of the input video, which will not affect the calculation of visual features according to this thread. So, the training should be fine and you can just ignore the warning.

Regarding your second question, we didn't count the exact number but there are not too many missing videos.

ayaan-together commented 3 months ago

Perfect!